Reference

The reference area is split by kind of thing, mirroring how the Kandra codebase itself is organized (see the platform's own CLAUDE.md for the authoritative layering):

• Attributes — the metadata vocabulary (Kandra.Attributes) that drives every generator.
• Entities (*Base) — the pure EF Core model layer.
• DTOs & Validators — the attribute-rich UI/API contract layer.
• Services — application/persistence services, behaviors, register writers/readers, posting, lookups, handlers, job scheduler.
• Source Generators — the Roslyn generators that remove hand-written plumbing.
• Register Engine — balance/turnover/info registers.
• Chart of Accounts / Posting — the independent accounting subsystem.
• Identity & Auth — users, roles, JWT, API keys, feature flags.

Two sources of truth

Reference content here comes from two places, and they're kept deliberately separate:

1. Hand-written pages — the overview.md (or similarly named) page at the top of each section above. These explain why a layer exists and how its pieces fit together — the kind of narrative context that doesn't belong in an XML doc comment. Edit these directly.

2. Generated API Reference — one page per type/member, produced mechanically from the C# /// XML doc comments in the source repos by scripts/sync-xml-docs.mjs. Never hand-edit anything under docs/reference/generated/ — edit the XML doc comment in the C# source and re-run the sync. See xml-docs/README.md for how the pipeline and its registry.csv change-tracking work.